Internal support

Internal Support functions. Not intended to be called direclty by a CMSIS-NN user.

Machine-readable model

• arm_nn_exp_poly_coeffs_f32attributePolynomial coefficients used by the float32 MVE exp approximation.
• arm_nn_exp2_lut_f32attributeLUT for 2^(i/256) used by the float32 LUT softmax approximation.
• arm_nn_softmax_floor_to_int_f32functionFloor of x as an int32t.
• arm_nn_softmax_fp32_from_bitsfunctionReinterpret a 32-bit pattern as a float32.
• arm_nn_softmax_exp2i_f32functionCompute 2^n as a float32 by building the exponent field directly.
• arm_nn_softmax_exp_taylor_f32functionTaylor/Estrin exp approximation for float32 softmax helpers.
• arm_nn_softmax_exp_lut_f32functionLUT-based exp approximation for float32 softmax helpers.
• arm_nn_softmax_exp_scalar_f32functionScalar exp approximation used by the float32 softmax paths.
• arm_nn_tanh_lut_f32attributeLUT for tanh(x) sampled over x in [0, 6] for float32 helpers.
• arm_memcpy_f32functionCopy a float32 vector.
• arm_memset_f32functionSet a float32 vector to a constant value.
• arm_nn_depthwise_conv1d_k3_nhwc_f32functionSpecialized NHWC depthwise 1D kernel for k=3, chmult=1 (float32).
• arm_nn_conv1d_k5_nhwc_f32functionSpecialized NHWC 1D convolution kernel for k=5 (float32).
• arm_nn_conv1d_k5_packed_f32functionSpecialized NHWC 1D convolution kernel for k=5 (float32, packed weights).
• arm_nn_conv1d_k3_nhwc_f32functionSpecialized NHWC 1D convolution kernel for k=3 (float32).
• arm_nn_conv1d_k3_packed_f32functionSpecialized NHWC 1D convolution kernel for k=3 (float32, packed weights).
• arm_nn_maxpool1d_k3s3_nhwc_f32functionSpecialized NHWC max-pool 1D kernel for k=3, s=3 (float32).
• arm_nn_maxpool1d_k2s2_nhwc_noclip_f32functionSpecialized NHWC max-pool 1D kernel for k=2, s=2 without output clamp (float32).
• arm_nn_maxpool1d_k2s2_nhwc_f32functionSpecialized NHWC max-pool 1D kernel for k=2, s=2 with clamp (float32).
• arm_nn_mat_mult_nt_t_f32functionMatrix multiply with non-transposed lhs and transposed rhs rows (float32).
• arm_nn_mat_mult_nt_n_packed_f32functionMatrix multiply with non-transposed lhs and packed non-transposed rhs (float32).
• arm_nn_pack_conv_patch_f32functionPack a single convolution patch into one row of a contiguous float32 patch matrix.
• arm_nn_softmax_1x2_f32functionSpecialized softmax helper for a single float32 row of length 2.
• ARM_NN_F16_ACC_BLOCKmacroBlockwise float16 accumulation on the MVE legs (AmbiqAI/ns-cmsis-nn#586).
• ARM_NN_F16_ACC_BLOCK_NONEmacro
• arm_nn_exp_poly_coeffs_f16attributePolynomial coefficients used by the float16 MVE exp approximation.
• arm_nn_exp2_lut_f16attributeQuantized binary16 LUT for 2^(i/256) used by float16 helpers.
• arm_nn_tanh_lut_f16attributeQuantized binary16 LUT for tanh(x) with x in [0, 4].
• arm_nn_softmax_fp16_from_bitsfunctionReinterpret a 16-bit pattern as a float16.
• arm_nn_softmax_floor_to_int_f16functionFloor of x as an int32t.
• arm_nn_softmax_exp2i_wide_f16functionCompute the power-of-two scale for the scalar float16 exp approximation.
• arm_nn_softmax_exp2i_f16functionCompute 2^n rounded to float16.
• arm_nn_softmax_exp_taylor_f16functionTaylor/Estrin exp approximation for float16 softmax helpers.
• arm_nn_softmax_exp_lut_f16functionLUT-based exp approximation for float16 softmax helpers.
• arm_nn_softmax_exp_scalar_f16functionScalar exp approximation used by the float16 softmax paths.
• arm_memcpy_f16functionCopy a float16 vector.
• arm_memset_f16functionSet a float16 vector to a constant value.
• arm_nn_depthwise_conv2x5_nhwc_f16functionSpecialized NHWC depthwise 2x5 kernel (float16).
• arm_nn_depthwise_conv1d_k3_nhwc_f16functionSpecialized NHWC depthwise 1D kernel for k=3, chmult=1 (float32).
• arm_nn_conv1d_k5_nhwc_f16functionSpecialized NHWC 1D convolution kernel for k=5 (float32).
• arm_nn_conv1d_k5_nhwc_f16_acc16functionSpecialized NHWC 1D convolution kernel for k=5 (float32).
• arm_nn_conv1d_k5_packed_f16functionSpecialized NHWC 1D convolution kernel for k=5 (float16, packed weights).
• arm_nn_conv1d_k5_packed_f16_acc16functionSpecialized NHWC 1D convolution kernel for k=5 (float16, packed weights).
• arm_nn_conv1d_k3_nhwc_f16functionSpecialized NHWC 1D convolution kernel for k=3 (float32).
• arm_nn_conv1d_k3_nhwc_f16_acc16functionSpecialized NHWC 1D convolution kernel for k=3 (float32).
• arm_nn_conv1d_k3_packed_f16functionSpecialized NHWC 1D convolution kernel for k=3 (float16, packed weights).
• arm_nn_conv1d_k3_packed_f16_acc16functionSpecialized NHWC 1D convolution kernel for k=3 (float16, packed weights).
• arm_nn_maxpool1d_k3s3_nhwc_f16functionSpecialized NHWC max-pool 1D kernel for k=3, s=3 (float16).
• arm_nn_maxpool1d_k2s2_nhwc_noclip_f16functionSpecialized NHWC max-pool 1D kernel for k=2, s=2 without output clamp (float16).
• arm_nn_maxpool1d_k2s2_nhwc_f16functionSpecialized NHWC max-pool 1D kernel for k=2, s=2 with clamp (float16).
• arm_nn_mat_mult_nt_t_f16functionMatrix multiply with non-transposed lhs and transposed rhs rows (float32).
• arm_nn_mat_mult_nt_t_f16_acc16functionarmnnmatmultnttf16 with every MVE accumulator lane in float16 (no blockwise fold).
• arm_nn_mat_mult_nt_n_packed_f16functionMatrix multiply with non-transposed lhs and packed non-transposed rhs (float16).
• arm_nn_mat_mult_nt_n_packed_f16_precisefunctionPacked float16 matrix multiply with shorter partial sums, for precise packed callers.
• arm_nn_mat_mult_nt_n_packed_f16_acc16functionarmnnmatmultntnpackedf16 with every MVE accumulator lane in float16 (no blockwise fold).
• arm_nn_lstm_step_f16functionUpdate LSTM function for an iteration step using float16 input, output and state.
• arm_nn_gru_step_f16functionUpdate GRU function for a single iteration step using float16 data.
• arm_nn_pack_conv_patch_f16functionPack a single convolution patch into one row of a contiguous float32 patch matrix.
• arm_nn_softmax_1x2_f16functionSpecialized softmax helper for a single float16 row of length 2.
• arm_nn_lstm_step_f32functionUpdate LSTM function for an iteration step using float32 input, output and state.
• arm_nn_gru_step_f32functionUpdate GRU function for a single iteration step using float32 data.

attribute

arm_nn_exp_poly_coeffs_f32

C

Polynomial coefficients used by the float32 MVE exp approximation.

Include/arm_nnsupportfunctions_flt.h:70

const float32_t arm_nn_exp_poly_coeffs_f32[8]

Polynomial coefficients used by the float32 MVE exp approximation.

attribute

arm_nn_exp2_lut_f32

C

LUT for 2^(i/256) used by the float32 LUT softmax approximation.

Include/arm_nnsupportfunctions_flt.h:78

const float32_t arm_nn_exp2_lut_f32[257]

LUT for 2^(i/256) used by the float32 LUT softmax approximation.

Stores 257 samples for i = 0..256 so interpolation can safely read lut[idx + 1] while indexing the 256 fractional segments.

function

arm_nn_softmax_floor_to_int_f32

C

Floor of x as an int32t.

Include/arm_nnsupportfunctions_flt.h:92

static int32_t arm_nn_softmax_floor_to_int_f32(float32_t x)

Floor of x as an int32_t.

Precondition: x must already be reduced to the int32_t range and must not be NaN the float-to-int conversion below is undefined otherwise. The only caller, arm_nn_softmax_exp_lut_f32(), guarantees this by clamping its input to [-80, 80] (NaN included, see there) before scaling by log2(e), which bounds x to +/-116.

Parameters

Parameters of arm_nn_softmax_floor_to_int_f32

Name	Type	Direction	Description
x	float32_t	in	Value to floor.

Returns

Returns of arm_nn_softmax_floor_to_int_f32

Description
Largest int32_t not greater than `x`.

function

arm_nn_softmax_fp32_from_bits

C

Reinterpret a 32-bit pattern as a float32.

Include/arm_nnsupportfunctions_flt.h:104

static float32_t arm_nn_softmax_fp32_from_bits(uint32_t bits)

Reinterpret a 32-bit pattern as a float32.

Parameters

Parameters of arm_nn_softmax_fp32_from_bits

Name	Type	Direction	Description
bits	uint32_t	in	IEEE-754 binary32 bit pattern.

Returns

Returns of arm_nn_softmax_fp32_from_bits

Description
The float32 value with the bit pattern `bits`.

function

arm_nn_softmax_exp2i_f32

C

Compute 2^n as a float32 by building the exponent field directly.

Include/arm_nnsupportfunctions_flt.h:121

static float32_t arm_nn_softmax_exp2i_f32(int32_t n)

Compute 2^n as a float32 by building the exponent field directly.

Parameters

Parameters of arm_nn_softmax_exp2i_f32

Name	Type	Direction	Description
n	int32_t	in	Integer exponent. Clamped to the normal float32 exponent range `[-126, 127]`.

Returns

Returns of arm_nn_softmax_exp2i_f32

Description
`2^n` as a float32.

function

arm_nn_softmax_exp_taylor_f32

C

Taylor/Estrin exp approximation for float32 softmax helpers.

Include/arm_nnsupportfunctions_flt.h:147

static float32_t arm_nn_softmax_exp_taylor_f32(float32_t x)

Taylor/Estrin exp approximation for float32 softmax helpers.

The polynomial is evaluated on r in [-ln2/2, ln2/2]. Coefficients come from the Maclaurin series of exp(r): exp(r) ~= 1 + r + r^2/2! + r^3/3! + r^4/4! + r^5/5! + r^6/6! Grouped via Estrin to reduce dependency depth: p = (1 + r) + r^2*(1/2 + r/6) + r^4*(1/24 + r/120) + r^6*(1/720)

Range reduction follows: x = n * ln(2) + r, exp(x) = exp(r) * 2^n

Parameters

Parameters of arm_nn_softmax_exp_taylor_f32

Name	Type	Direction	Description
x	float32_t	in	Exponent argument. Clamped to `[-80, 80]` before evaluation.

Returns

Returns of arm_nn_softmax_exp_taylor_f32

Description
Approximation of `exp(x)`.

function

arm_nn_softmax_exp_lut_f32

C

LUT-based exp approximation for float32 softmax helpers.

Include/arm_nnsupportfunctions_flt.h:184

static float32_t arm_nn_softmax_exp_lut_f32(float32_t x)

LUT-based exp approximation for float32 softmax helpers.

Splits x * log2(e) into an integer part handled by arm_nn_softmax_exp2i_f32() and a fractional part interpolated linearly from arm_nn_exp2_lut_f32.

Parameters

Parameters of arm_nn_softmax_exp_lut_f32

Name	Type	Direction	Description
x	float32_t	in	Exponent argument. Clamped to `[-80, 80]` before evaluation; NaN is flushed to `80`.

Returns

Returns of arm_nn_softmax_exp_lut_f32

Description
Approximation of `exp(x)`.

function

arm_nn_softmax_exp_scalar_f32

C

Scalar exp approximation used by the float32 softmax paths.

Include/arm_nnsupportfunctions_flt.h:254

static float32_t arm_nn_softmax_exp_scalar_f32(float32_t x)

Scalar exp approximation used by the float32 softmax paths.

Dispatches to arm_nn_softmax_exp_taylor_f32() when ARM_NN_USE_EXP_TAYLOR is defined and to arm_nn_softmax_exp_lut_f32() otherwise.

Parameters

Parameters of arm_nn_softmax_exp_scalar_f32

Name	Type	Direction	Description
x	float32_t	in	Exponent argument.

Returns

Returns of arm_nn_softmax_exp_scalar_f32

Description
Approximation of `exp(x)`.

attribute

arm_nn_tanh_lut_f32

C

LUT for tanh(x) sampled over x in [0, 6] for float32 helpers.

Include/arm_nnsupportfunctions_flt.h:276

const float32_t arm_nn_tanh_lut_f32[385]

LUT for tanh(x) sampled over x in [0, 6] for float32 helpers.

Stores 385 samples so interpolation can safely read lut[idx + 1] while indexing the 384 fractional segments across the interval. The grid spacing (6/384 == 1/64) matches the earlier 257-entry [0, 4] table, so entries 0..256 are bit-identical to it and the index multiplier is unchanged. Generated by scripts/gen_tanh_lut_f32.py.

function

arm_memcpy_f32

C

Copy a float32 vector.

Include/arm_nnsupportfunctions_flt.h:311

static void arm_memcpy_f32(float32_t *dst, const float32_t *src, uint32_t block_size)

Copy a float32 vector.

Parameters

Parameters of arm_memcpy_f32

Name	Type	Direction	Description
dst	float32_t *	out	Destination buffer.
src	const float32_t *	in	Source buffer.
block_size	uint32_t	in	Number of elements to copy.

function

arm_memset_f32

C

Set a float32 vector to a constant value.

Include/arm_nnsupportfunctions_flt.h:334

static void arm_memset_f32(float32_t *dst, const float32_t val, uint32_t block_size)

Set a float32 vector to a constant value.

Parameters

Parameters of arm_memset_f32

Name	Type	Direction	Description
dst	float32_t *	out	Destination buffer.
val	const float32_t	in	Fill value.
block_size	uint32_t	in	Number of elements to write.

function

arm_nn_depthwise_conv1d_k3_nhwc_f32

C

Specialized NHWC depthwise 1D kernel for k=3, chmult=1 (float32).

Include/arm_nnsupportfunctions_flt.h:367

void arm_nn_depthwise_conv1d_k3_nhwc_f32(
    const float32_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    const float32_t *kernel,
    const float32_t *b,
    float32_t *out,
    int32_t out_w
)

Specialized NHWC depthwise 1D kernel for k=3, ch_mult=1 (float32).

Parameters

Parameters of arm_nn_depthwise_conv1d_k3_nhwc_f32

Name	Type	Direction	Description
x_nhwc	const float32_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of input (and output) channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
kernel	const float32_t *	in	Depthwise weights with shape `[3][in_c]`.
b	const float32_t *	in	Optional bias vector of `in_c` elements. May be NULL.
out	float32_t *	out	Output row in NHWC layout with shape `[out_w][in_c]`.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `ow..ow+2`.

function

arm_nn_conv1d_k5_nhwc_f32

C

Specialized NHWC 1D convolution kernel for k=5 (float32).

Include/arm_nnsupportfunctions_flt.h:387

void arm_nn_conv1d_k5_nhwc_f32(
    const float32_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    const float32_t *kernel,
    const float32_t *b,
    float32_t *out,
    int32_t out_c,
    int32_t out_w
)

Specialized NHWC 1D convolution kernel for k=5 (float32).

Parameters

Parameters of arm_nn_conv1d_k5_nhwc_f32

Name	Type	Direction	Description
x_nhwc	const float32_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of input channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
kernel	const float32_t *	in	Weights with shape `[out_c][5][in_c]`.
b	const float32_t *	in	Optional bias vector of `out_c` elements. May be NULL.
out	float32_t *	out	Output row in NHWC layout with shape `[out_w][out_c]`.
out_c	int32_t	in	Number of output channels.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `ow..ow+4`.

function

arm_nn_conv1d_k5_packed_f32

C

Specialized NHWC 1D convolution kernel for k=5 (float32, packed weights).

Include/arm_nnsupportfunctions_flt.h:411

void arm_nn_conv1d_k5_packed_f32(
    const float32_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    const float32_t *kernel_packed,
    const float32_t *b,
    float32_t *out,
    int32_t out_c,
    int32_t out_w
)

Specialized NHWC 1D convolution kernel for k=5 (float32, packed weights).

The packed kernel uses the same NTxN RHS layout as arm_nn_mat_mult_nt_n_packed_f32, i.e. [(5 * in_c)][out_c_block_of_4].

Parameters

Parameters of arm_nn_conv1d_k5_packed_f32

Name	Type	Direction	Description
x_nhwc	const float32_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of input channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
kernel_packed	const float32_t *	in	Weights packed in output-channel blocks of 4 as described above.
b	const float32_t *	in	Optional bias vector of `out_c` elements. May be NULL.
out	float32_t *	out	Output row in NHWC layout with shape `[out_w][out_c]`.
out_c	int32_t	in	Number of output channels.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `ow..ow+4`.

function

arm_nn_conv1d_k3_nhwc_f32

C

Specialized NHWC 1D convolution kernel for k=3 (float32).

Include/arm_nnsupportfunctions_flt.h:432

void arm_nn_conv1d_k3_nhwc_f32(
    const float32_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    const float32_t *kernel,
    const float32_t *b,
    float32_t *out,
    int32_t out_c,
    int32_t out_w
)

Specialized NHWC 1D convolution kernel for k=3 (float32).

Parameters

Parameters of arm_nn_conv1d_k3_nhwc_f32

Name	Type	Direction	Description
x_nhwc	const float32_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of input channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
kernel	const float32_t *	in	Weights with shape `[out_c][3][in_c]`.
b	const float32_t *	in	Optional bias vector of `out_c` elements. May be NULL.
out	float32_t *	out	Output row in NHWC layout with shape `[out_w][out_c]`.
out_c	int32_t	in	Number of output channels.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `ow..ow+2`.

function

arm_nn_conv1d_k3_packed_f32

C

Specialized NHWC 1D convolution kernel for k=3 (float32, packed weights).

Include/arm_nnsupportfunctions_flt.h:456

void arm_nn_conv1d_k3_packed_f32(
    const float32_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    const float32_t *kernel_packed,
    const float32_t *b,
    float32_t *out,
    int32_t out_c,
    int32_t out_w
)

Specialized NHWC 1D convolution kernel for k=3 (float32, packed weights).

The packed kernel uses the same NTxN RHS layout as arm_nn_mat_mult_nt_n_packed_f32, i.e. [(3 * in_c)][out_c_block_of_4].

Parameters

Parameters of arm_nn_conv1d_k3_packed_f32

Name	Type	Direction	Description
x_nhwc	const float32_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of input channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
kernel_packed	const float32_t *	in	Weights packed in output-channel blocks of 4 as described above.
b	const float32_t *	in	Optional bias vector of `out_c` elements. May be NULL.
out	float32_t *	out	Output row in NHWC layout with shape `[out_w][out_c]`.
out_c	int32_t	in	Number of output channels.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `ow..ow+2`.

function

arm_nn_maxpool1d_k3s3_nhwc_f32

C

Specialized NHWC max-pool 1D kernel for k=3, s=3 (float32).

Include/arm_nnsupportfunctions_flt.h:474

void arm_nn_maxpool1d_k3s3_nhwc_f32(const float32_t *x_nhwc, int32_t in_c, int32_t in_w, float32_t *out, int32_t out_w)

Specialized NHWC max-pool 1D kernel for k=3, s=3 (float32).

Parameters

Parameters of arm_nn_maxpool1d_k3s3_nhwc_f32

Name	Type	Direction	Description
x_nhwc	const float32_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
out	float32_t *	out	Output row in NHWC layout with shape `[out_w][in_c]`.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `3*ow..3*ow+2`.

function

arm_nn_maxpool1d_k2s2_nhwc_noclip_f32

C

Specialized NHWC max-pool 1D kernel for k=2, s=2 without output clamp (float32).

Include/arm_nnsupportfunctions_flt.h:489

void arm_nn_maxpool1d_k2s2_nhwc_noclip_f32(
    const float32_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    float32_t *out,
    int32_t out_w
)

Specialized NHWC max-pool 1D kernel for k=2, s=2 without output clamp (float32).

Parameters

Parameters of arm_nn_maxpool1d_k2s2_nhwc_noclip_f32

Name	Type	Direction	Description
x_nhwc	const float32_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
out	float32_t *	out	Output row in NHWC layout with shape `[out_w][in_c]`.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `2*ow..2*ow+1`.

function

arm_nn_maxpool1d_k2s2_nhwc_f32

C

Specialized NHWC max-pool 1D kernel for k=2, s=2 with clamp (float32).

Include/arm_nnsupportfunctions_flt.h:506

void arm_nn_maxpool1d_k2s2_nhwc_f32(
    const float32_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    float32_t *out,
    int32_t out_w,
    float32_t act_min,
    float32_t act_max
)

Specialized NHWC max-pool 1D kernel for k=2, s=2 with clamp (float32).

Parameters

Parameters of arm_nn_maxpool1d_k2s2_nhwc_f32

Name	Type	Direction	Description
x_nhwc	const float32_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
out	float32_t *	out	Output row in NHWC layout with shape `[out_w][in_c]`.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `2*ow..2*ow+1`.
act_min	float32_t	in	Lower clamp bound applied to `out`.
act_max	float32_t	in	Upper clamp bound applied to `out`.

function

arm_nn_mat_mult_nt_t_f32

C

Matrix multiply with non-transposed lhs and transposed rhs rows (float32).

Include/arm_nnsupportfunctions_flt.h:529

arm_cmsis_nn_status arm_nn_mat_mult_nt_t_f32(
    const float32_t *lhs,
    const float32_t *rhs,
    const float32_t *bias,
    float32_t *dst,
    int32_t lhs_rows,
    int32_t rhs_rows,
    int32_t rhs_cols,
    int32_t row_address_offset,
    float32_t activation_min,
    float32_t activation_max
)

Matrix multiply with non-transposed lhs and transposed rhs rows (float32).

Parameters

Parameters of arm_nn_mat_mult_nt_t_f32

Name	Type	Direction	Description
lhs	const float32_t *	in	Left-hand matrix stored row-major.
rhs	const float32_t *	in	Right-hand matrix stored row-major, one row per output channel.
bias	const float32_t *	in	Optional bias vector.
dst	float32_t *	out	Output matrix.
lhs_rows	int32_t	in	Number of rows in `lhs`.
rhs_rows	int32_t	in	Number of rows in `rhs`.
rhs_cols	int32_t	in	Number of columns in `rhs`.
row_address_offset	int32_t	in	Output row stride, expressed in elements.
activation_min	float32_t	in	Lower clamp bound.
activation_max	float32_t	in	Upper clamp bound.

Returns

Returns of arm_nn_mat_mult_nt_t_f32

Description
`ARM_CMSIS_NN_SUCCESS` on success or `ARM_CMSIS_NN_ARG_ERROR` on invalid arguments.

function

arm_nn_mat_mult_nt_n_packed_f32

C

Matrix multiply with non-transposed lhs and packed non-transposed rhs (float32).

Include/arm_nnsupportfunctions_flt.h:556

arm_cmsis_nn_status arm_nn_mat_mult_nt_n_packed_f32(
    const float32_t *lhs,
    const float32_t *rhs_packed,
    const float32_t *bias,
    float32_t *dst,
    int32_t lhs_rows,
    int32_t rhs_rows,
    int32_t rhs_cols,
    int32_t row_address_offset,
    float32_t activation_min,
    float32_t activation_max
)

Matrix multiply with non-transposed lhs and packed non-transposed rhs (float32).

Parameters

Parameters of arm_nn_mat_mult_nt_n_packed_f32

Name	Type	Direction	Description
lhs	const float32_t *	in	Left-hand matrix stored row-major with logical shape `[lhs_rows, rhs_cols]`.
rhs_packed	const float32_t *	in	Right-hand matrix with logical shape `[rhs_cols, rhs_rows]`, packed in column blocks of 4. The final block uses the same packed stride and inactive tail lanes are ignored.
bias	const float32_t *	in	Optional bias vector.
dst	float32_t *	out	Output matrix.
lhs_rows	int32_t	in	Number of rows in `lhs`.
rhs_rows	int32_t	in	Number of logical output columns in the unpacked rhs matrix.
rhs_cols	int32_t	in	Shared reduction dimension `K`.
row_address_offset	int32_t	in	Output row stride, expressed in elements.
activation_min	float32_t	in	Lower clamp bound.
activation_max	float32_t	in	Upper clamp bound.

Returns

Returns of arm_nn_mat_mult_nt_n_packed_f32

Description
`ARM_CMSIS_NN_SUCCESS` on success or `ARM_CMSIS_NN_ARG_ERROR` on invalid arguments.

function

arm_nn_pack_conv_patch_f32

C

Pack a single convolution patch into one row of a contiguous float32 patch matrix.

Include/arm_nnsupportfunctions_flt.h:590

void arm_nn_pack_conv_patch_f32(
    const float32_t *input,
    int32_t in_h,
    int32_t in_w,
    int32_t in_c,
    int32_t kernel_h,
    int32_t kernel_w,
    int32_t stride_h,
    int32_t stride_w,
    int32_t pad_h,
    int32_t pad_w,
    int32_t dilation_h,
    int32_t dilation_w,
    int32_t out_y,
    int32_t out_x,
    float32_t pad_value,
    float32_t *patch_row
)

Pack a single convolution patch into one row of a contiguous float32 patch matrix.

Developers familiar with im2row/im2col terminology can think of this as packing one output patch into one row.

Parameters

Parameters of arm_nn_pack_conv_patch_f32

Name	Type	Direction	Description
input	const float32_t *	in	Input tensor for one batch in NHWC layout with shape `[in_h][in_w][in_c]`.
in_h	int32_t	in	Input height.
in_w	int32_t	in	Input width.
in_c	int32_t	in	Number of input channels.
kernel_h	int32_t	in	Kernel height.
kernel_w	int32_t	in	Kernel width.
stride_h	int32_t	in	Vertical stride.
stride_w	int32_t	in	Horizontal stride.
pad_h	int32_t	in	Top padding.
pad_w	int32_t	in	Left padding.
dilation_h	int32_t	in	Vertical dilation.
dilation_w	int32_t	in	Horizontal dilation.
out_y	int32_t	in	Output row index of the patch to pack.
out_x	int32_t	in	Output column index of the patch to pack.
pad_value	float32_t	in	Value written for taps that fall outside the input.
patch_row	float32_t *	out	Destination row of `kernel_h * kernel_w * in_c` elements, ordered `[kernel_h][kernel_w][in_c]`.

function

arm_nn_softmax_1x2_f32

C

Specialized softmax helper for a single float32 row of length 2.

Include/arm_nnsupportfunctions_flt.h:613

void arm_nn_softmax_1x2_f32(const float32_t *in, float32_t *out)

Specialized softmax helper for a single float32 row of length 2.

Parameters

Parameters of arm_nn_softmax_1x2_f32

Name	Type	Direction	Description
in	const float32_t *	in	Pointer to two contiguous float32 input values.
out	float32_t *	out	Pointer to two contiguous float32 output values.

macro

ARM_NN_F16_ACC_BLOCK

C

Blockwise float16 accumulation on the MVE legs (AmbiqAI/ns-cmsis-nn#586).

Include/arm_nnsupportfunctions_flt.h:626

#define ARM_NN_F16_ACC_BLOCK (32)

Blockwise float16 accumulation on the MVE legs (AmbiqAI/ns-cmsis-nn#586).

A float16 accumulator lane sums at most ARM_NN_F16_ACC_BLOCK taps, in the kernel’s tap order, before its partial is widened exactly and added into a float32 accumulator; the float32 sum rounds to float16 once. The _acc16 entries instantiate the same kernel bodies with ARM_NN_F16_ACC_BLOCK_NONE, which never folds.

macro

ARM_NN_F16_ACC_BLOCK_NONE

C

Include/arm_nnsupportfunctions_flt.h:627

#define ARM_NN_F16_ACC_BLOCK_NONE (INT32_MAX)

attribute

arm_nn_exp_poly_coeffs_f16

C

Polynomial coefficients used by the float16 MVE exp approximation.

Include/arm_nnsupportfunctions_flt.h:636

const float32_t arm_nn_exp_poly_coeffs_f16[8]

Polynomial coefficients used by the float16 MVE exp approximation.

The float16 MVE helper evaluates the polynomial in widened float32 lanes, but it uses a dedicated coefficient table to keep the float16 path isolated from the float32 feature gate and softmax support stack.

attribute

arm_nn_exp2_lut_f16

C

Quantized binary16 LUT for 2^(i/256) used by float16 helpers.

Include/arm_nnsupportfunctions_flt.h:644

const uint16_t arm_nn_exp2_lut_f16[257]

Quantized binary16 LUT for 2^(i/256) used by float16 helpers.

Stores 257 samples for i = 0..256 so interpolation can safely read lut[idx + 1] while indexing the 256 fractional segments.

attribute

arm_nn_tanh_lut_f16

C

Quantized binary16 LUT for tanh(x) with x in [0, 4].

Include/arm_nnsupportfunctions_flt.h:652

const uint16_t arm_nn_tanh_lut_f16[257]

Quantized binary16 LUT for tanh(x) with x in [0, 4].

Stores 257 samples so interpolation can safely read lut[idx + 1] while indexing the 256 fractional segments across the interval.

function

arm_nn_softmax_fp16_from_bits

C

Reinterpret a 16-bit pattern as a float16.

Include/arm_nnsupportfunctions_flt.h:660

static float16_t arm_nn_softmax_fp16_from_bits(uint16_t bits)

Reinterpret a 16-bit pattern as a float16.

Parameters

Parameters of arm_nn_softmax_fp16_from_bits

Name	Type	Direction	Description
bits	uint16_t	in	IEEE-754 binary16 bit pattern.

Returns

Returns of arm_nn_softmax_fp16_from_bits

Description
The float16 value with the bit pattern `bits`.

function

arm_nn_softmax_floor_to_int_f16

C

Floor of x as an int32t.

Include/arm_nnsupportfunctions_flt.h:677

static int32_t arm_nn_softmax_floor_to_int_f16(float16_t x)

Floor of x as an int32_t.

Parameters

Parameters of arm_nn_softmax_floor_to_int_f16

Name	Type	Direction	Description
x	float16_t	in	Value to floor. Must be finite and within the int32_t range.

Returns

Returns of arm_nn_softmax_floor_to_int_f16

Description
Largest int32_t not greater than `x`.

function

arm_nn_softmax_exp2i_wide_f16

C

Compute the power-of-two scale for the scalar float16 exp approximation.

Include/arm_nnsupportfunctions_flt.h:694

static float32_t arm_nn_softmax_exp2i_wide_f16(int32_t n)

Compute the power-of-two scale for the scalar float16 exp approximation.

Keep the scale in float32 until the approximation is rounded to float16. Rounding the scale first would lose subnormal results when the fractional factor lifts a value above the float16 underflow threshold. The upper clamp is unchanged.

Parameters

Parameters of arm_nn_softmax_exp2i_wide_f16

Name	Type	Direction	Description
n	int32_t	in	Integer exponent. Clamped to `[-126, 15]`.

Returns

Returns of arm_nn_softmax_exp2i_wide_f16

Description
`2^n` as a float32 intermediate.

function

arm_nn_softmax_exp2i_f16

C

Compute 2^n rounded to float16.

Include/arm_nnsupportfunctions_flt.h:712

static float16_t arm_nn_softmax_exp2i_f16(int32_t n)

Compute 2^n rounded to float16.

Parameters

Parameters of arm_nn_softmax_exp2i_f16

Name	Type	Direction	Description
n	int32_t	in	Integer exponent. Clamped above at 15; small results round to subnormal or zero.

Returns

Returns of arm_nn_softmax_exp2i_f16

Description
`2^n` as a float16.

function

arm_nn_softmax_exp_taylor_f16

C

Taylor/Estrin exp approximation for float16 softmax helpers.

Include/arm_nnsupportfunctions_flt.h:723

static float16_t arm_nn_softmax_exp_taylor_f16(float16_t x)

Taylor/Estrin exp approximation for float16 softmax helpers.

The evaluation uses float32 intermediates to keep the approximation stable, but it is fully independent from the float32 softmax support tables.

Parameters

Parameters of arm_nn_softmax_exp_taylor_f16

Name	Type	Direction	Description
x	float16_t	in	Exponent argument. Clamped to `[-80, 80]` before evaluation.

Returns

Returns of arm_nn_softmax_exp_taylor_f16

Description
Approximation of `exp(x)`.

function

arm_nn_softmax_exp_lut_f16

C

LUT-based exp approximation for float16 softmax helpers.

Include/arm_nnsupportfunctions_flt.h:759

static float16_t arm_nn_softmax_exp_lut_f16(float16_t x)

LUT-based exp approximation for float16 softmax helpers.

Splits x * log2(e) into an integer part handled by arm_nn_softmax_exp2i_wide_f16() and a fractional part interpolated linearly from arm_nn_exp2_lut_f16, using float32 intermediates.

Parameters

Parameters of arm_nn_softmax_exp_lut_f16

Name	Type	Direction	Description
x	float16_t	in	Exponent argument. Clamped to `[-80, 80]` before evaluation.

Returns

Returns of arm_nn_softmax_exp_lut_f16

Description
Approximation of `exp(x)`.

function

arm_nn_softmax_exp_scalar_f16

C

Scalar exp approximation used by the float16 softmax paths.

Include/arm_nnsupportfunctions_flt.h:807

static float16_t arm_nn_softmax_exp_scalar_f16(float16_t x)

Scalar exp approximation used by the float16 softmax paths.

Dispatches to arm_nn_softmax_exp_taylor_f16() when ARM_NN_USE_EXP_TAYLOR is defined and to arm_nn_softmax_exp_lut_f16() otherwise.

Parameters

Parameters of arm_nn_softmax_exp_scalar_f16

Name	Type	Direction	Description
x	float16_t	in	Exponent argument.

Returns

Returns of arm_nn_softmax_exp_scalar_f16

Description
Approximation of `exp(x)`.

function

arm_memcpy_f16

C

Copy a float16 vector.

Include/arm_nnsupportfunctions_flt.h:999

static void arm_memcpy_f16(float16_t *dst, const float16_t *src, uint32_t block_size)

Copy a float16 vector.

Parameters

Parameters of arm_memcpy_f16

Name	Type	Direction	Description
dst	float16_t *	out	Destination buffer.
src	const float16_t *	in	Source buffer.
block_size	uint32_t	in	Number of elements to copy.

function

arm_memset_f16

C

Set a float16 vector to a constant value.

Include/arm_nnsupportfunctions_flt.h:1022

static void arm_memset_f16(float16_t *dst, const float16_t val, uint32_t block_size)

Set a float16 vector to a constant value.

Parameters

Parameters of arm_memset_f16

Name	Type	Direction	Description
dst	float16_t *	out	Destination buffer.
val	const float16_t	in	Fill value.
block_size	uint32_t	in	Number of elements to write.

function

arm_nn_depthwise_conv2x5_nhwc_f16

C

Specialized NHWC depthwise 2x5 kernel (float16).

Include/arm_nnsupportfunctions_flt.h:1059

void arm_nn_depthwise_conv2x5_nhwc_f16(
    const float16_t *x_nhwc,
    int32_t batches,
    int32_t in_c,
    int32_t in_w,
    int32_t ch_mult,
    const float16_t *kernel,
    const float16_t *b,
    float16_t *out,
    int32_t out_w,
    float16_t act_min,
    float16_t act_max
)

Specialized NHWC depthwise 2x5 kernel (float16).

Parameters

Parameters of arm_nn_depthwise_conv2x5_nhwc_f16

Name	Type	Direction	Description
x_nhwc	const float16_t *	in	Input tensor in NHWC layout with shape `[batches][2][in_w][in_c]`.
batches	int32_t	in	Number of batches.
in_c	int32_t	in	Number of input channels.
in_w	int32_t	in	Input width.
ch_mult	int32_t	in	Channel multiplier; the output has `in_c * ch_mult` channels.
kernel	const float16_t *	in	Depthwise weights with shape `[2][5][in_c * ch_mult]`.
b	const float16_t *	in	Optional bias vector of `in_c * ch_mult` elements. May be NULL.
out	float16_t *	out	Output tensor in NHWC layout with shape `[batches][1][out_w][in_c * ch_mult]`.
out_w	int32_t	in	Output width. Output position `ow` reads input columns `ow..ow+4`.
act_min	float16_t	in	Lower clamp bound applied to `out`.
act_max	float16_t	in	Upper clamp bound applied to `out`.

function

arm_nn_depthwise_conv1d_k3_nhwc_f16

C

Specialized NHWC depthwise 1D kernel for k=3, chmult=1 (float32).

Include/arm_nnsupportfunctions_flt.h:1074

void arm_nn_depthwise_conv1d_k3_nhwc_f16(
    const float16_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    const float16_t *kernel,
    const float16_t *b,
    float16_t *out,
    int32_t out_w
)

Specialized NHWC depthwise 1D kernel for k=3, ch_mult=1 (float32).

Parameters

Parameters of arm_nn_depthwise_conv1d_k3_nhwc_f16

Name	Type	Direction	Description
x_nhwc	const float16_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of input (and output) channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
kernel	const float16_t *	in	Depthwise weights with shape `[3][in_c]`.
b	const float16_t *	in	Optional bias vector of `in_c` elements. May be NULL.
out	float16_t *	out	Output row in NHWC layout with shape `[out_w][in_c]`.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `ow..ow+2`.

function

arm_nn_conv1d_k5_nhwc_f16

C

Specialized NHWC 1D convolution kernel for k=5 (float32).

Include/arm_nnsupportfunctions_flt.h:1093

void arm_nn_conv1d_k5_nhwc_f16(
    const float16_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    const float16_t *kernel,
    const float16_t *b,
    float16_t *out,
    int32_t out_c,
    int32_t out_w
)

Specialized NHWC 1D convolution kernel for k=5 (float32).

Note

MVE leg: blockwise float16 accumulation (AmbiqAI/ns-cmsis-nn#586). Input channel c feeds lane c % 8 with 5 taps per channel step; above 32 taps per output, a lane’s float16 partial covers at most 6 channel steps; each block’s lanes are folded into float32 pair accumulators (arm_nn_f16_fold_pairs_f32), which are summed once (arm_nn_f16_pairs_sum_f32), the bias is added in float32 and the total rounds to float16 once. Up to 32 taps: float16 lanes, a float16 reduction and the bias added in float16, as before, in the order the compiler gives them (it may reorder them under -ffast-math); only the fold’s order is fixed. The scalar leg accumulates in float32 (#449, #465).

Parameters

Parameters of arm_nn_conv1d_k5_nhwc_f16

Name	Type	Direction	Description
x_nhwc	const float16_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of input channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
kernel	const float16_t *	in	Weights with shape `[out_c][5][in_c]`.
b	const float16_t *	in	Optional bias vector of `out_c` elements. May be NULL.
out	float16_t *	out	Output row in NHWC layout with shape `[out_w][out_c]`.
out_c	int32_t	in	Number of output channels.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `ow..ow+4`.

function

arm_nn_conv1d_k5_nhwc_f16_acc16

C

Specialized NHWC 1D convolution kernel for k=5 (float32).

Include/arm_nnsupportfunctions_flt.h:1108

void arm_nn_conv1d_k5_nhwc_f16_acc16(
    const float16_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    const float16_t *kernel,
    const float16_t *b,
    float16_t *out,
    int32_t out_c,
    int32_t out_w
)

Specialized NHWC 1D convolution kernel for k=5 (float32).

Note

MVE leg: blockwise float16 accumulation (AmbiqAI/ns-cmsis-nn#586). Input channel c feeds lane c % 8 with 5 taps per channel step; above 32 taps per output, a lane’s float16 partial covers at most 6 channel steps; each block’s lanes are folded into float32 pair accumulators (arm_nn_f16_fold_pairs_f32), which are summed once (arm_nn_f16_pairs_sum_f32), the bias is added in float32 and the total rounds to float16 once. Up to 32 taps: float16 lanes, a float16 reduction and the bias added in float16, as before, in the order the compiler gives them (it may reorder them under -ffast-math); only the fold’s order is fixed. The scalar leg accumulates in float32 (#449, #465).

Note

Every MVE accumulator lane stays in float16 (no blockwise fold, AmbiqAI/ns-cmsis-nn#586); the scalar leg is the same as arm_nn_conv1d_k5_nhwc_f16.

Parameters

Parameters of arm_nn_conv1d_k5_nhwc_f16_acc16

Name	Type	Direction	Description
x_nhwc	const float16_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of input channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
kernel	const float16_t *	in	Weights with shape `[out_c][5][in_c]`.
b	const float16_t *	in	Optional bias vector of `out_c` elements. May be NULL.
out	float16_t *	out	Output row in NHWC layout with shape `[out_w][out_c]`.
out_c	int32_t	in	Number of output channels.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `ow..ow+4`.

function

arm_nn_conv1d_k5_packed_f16

C

Specialized NHWC 1D convolution kernel for k=5 (float16, packed weights).

Include/arm_nnsupportfunctions_flt.h:1138

void arm_nn_conv1d_k5_packed_f16(
    const float16_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    const float16_t *kernel_packed,
    const float16_t *b,
    float16_t *out,
    int32_t out_c,
    int32_t out_w
)

Specialized NHWC 1D convolution kernel for k=5 (float16, packed weights).

The packed kernel uses the same NTxN RHS layout as arm_nn_mat_mult_nt_n_packed_f16, i.e. [(5 * in_c)][out_c_block_of_8].

Note

Accumulation width per leg. Scalar leg (non-MVE builds and ARM_MATH_AUTOVECTORIZE): bias and products accumulate in float32 and round to float16 once at the store (AmbiqAI/ns-cmsis-nn#449, #465). MVE leg: blockwise (#586): a lane’s float16 partial covers at most 6 input channels (30 taps, bias first) before it is widened into a float32 accumulator; one rounding at the store. Up to 32 taps (in_c <= 6) this is the float16-lane result.

Parameters

Parameters of arm_nn_conv1d_k5_packed_f16

Name	Type	Direction	Description
x_nhwc	const float16_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of input channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
kernel_packed	const float16_t *	in	Weights packed in output-channel blocks of 8 as described above.
b	const float16_t *	in	Optional bias vector of `out_c` elements. May be NULL.
out	float16_t *	out	Output row in NHWC layout with shape `[out_w][out_c]`.
out_c	int32_t	in	Number of output channels.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `ow..ow+4`.

function

arm_nn_conv1d_k5_packed_f16_acc16

C

Specialized NHWC 1D convolution kernel for k=5 (float16, packed weights).

Include/arm_nnsupportfunctions_flt.h:1153

void arm_nn_conv1d_k5_packed_f16_acc16(
    const float16_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    const float16_t *kernel_packed,
    const float16_t *b,
    float16_t *out,
    int32_t out_c,
    int32_t out_w
)

Specialized NHWC 1D convolution kernel for k=5 (float16, packed weights).

The packed kernel uses the same NTxN RHS layout as arm_nn_mat_mult_nt_n_packed_f16, i.e. [(5 * in_c)][out_c_block_of_8].

Note

Accumulation width per leg. Scalar leg (non-MVE builds and ARM_MATH_AUTOVECTORIZE): bias and products accumulate in float32 and round to float16 once at the store (AmbiqAI/ns-cmsis-nn#449, #465). MVE leg: blockwise (#586): a lane’s float16 partial covers at most 6 input channels (30 taps, bias first) before it is widened into a float32 accumulator; one rounding at the store. Up to 32 taps (in_c <= 6) this is the float16-lane result.

Note

Every MVE accumulator lane stays in float16 (no blockwise fold, AmbiqAI/ns-cmsis-nn#586); the scalar leg is the same as arm_nn_conv1d_k5_packed_f16.

Parameters

Parameters of arm_nn_conv1d_k5_packed_f16_acc16

Name	Type	Direction	Description
x_nhwc	const float16_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of input channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
kernel_packed	const float16_t *	in	Weights packed in output-channel blocks of 8 as described above.
b	const float16_t *	in	Optional bias vector of `out_c` elements. May be NULL.
out	float16_t *	out	Output row in NHWC layout with shape `[out_w][out_c]`.
out_c	int32_t	in	Number of output channels.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `ow..ow+4`.

function

arm_nn_conv1d_k3_nhwc_f16

C

Specialized NHWC 1D convolution kernel for k=3 (float32).

Include/arm_nnsupportfunctions_flt.h:1173

void arm_nn_conv1d_k3_nhwc_f16(
    const float16_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    const float16_t *kernel,
    const float16_t *b,
    float16_t *out,
    int32_t out_c,
    int32_t out_w
)

Specialized NHWC 1D convolution kernel for k=3 (float32).

Note

MVE leg: blockwise float16 accumulation (AmbiqAI/ns-cmsis-nn#586). Input channel c feeds lane c % 8 with 3 taps per channel step; above 32 taps per output, a lane’s float16 partial covers at most 10 channel steps; each block’s lanes are folded into float32 pair accumulators (arm_nn_f16_fold_pairs_f32), which are summed once (arm_nn_f16_pairs_sum_f32), the bias is added in float32 and the total rounds to float16 once. Up to 32 taps: float16 lanes, a float16 reduction and the bias added in float16, as before, in the order the compiler gives them (it may reorder them under -ffast-math); only the fold’s order is fixed. The scalar leg accumulates in float32 (#449, #465).

Parameters

Parameters of arm_nn_conv1d_k3_nhwc_f16

Name	Type	Direction	Description
x_nhwc	const float16_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of input channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
kernel	const float16_t *	in	Weights with shape `[out_c][3][in_c]`.
b	const float16_t *	in	Optional bias vector of `out_c` elements. May be NULL.
out	float16_t *	out	Output row in NHWC layout with shape `[out_w][out_c]`.
out_c	int32_t	in	Number of output channels.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `ow..ow+2`.

function

arm_nn_conv1d_k3_nhwc_f16_acc16

C

Specialized NHWC 1D convolution kernel for k=3 (float32).

Include/arm_nnsupportfunctions_flt.h:1188

void arm_nn_conv1d_k3_nhwc_f16_acc16(
    const float16_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    const float16_t *kernel,
    const float16_t *b,
    float16_t *out,
    int32_t out_c,
    int32_t out_w
)

Specialized NHWC 1D convolution kernel for k=3 (float32).

Note

MVE leg: blockwise float16 accumulation (AmbiqAI/ns-cmsis-nn#586). Input channel c feeds lane c % 8 with 3 taps per channel step; above 32 taps per output, a lane’s float16 partial covers at most 10 channel steps; each block’s lanes are folded into float32 pair accumulators (arm_nn_f16_fold_pairs_f32), which are summed once (arm_nn_f16_pairs_sum_f32), the bias is added in float32 and the total rounds to float16 once. Up to 32 taps: float16 lanes, a float16 reduction and the bias added in float16, as before, in the order the compiler gives them (it may reorder them under -ffast-math); only the fold’s order is fixed. The scalar leg accumulates in float32 (#449, #465).

Note

Every MVE accumulator lane stays in float16 (no blockwise fold, AmbiqAI/ns-cmsis-nn#586); the scalar leg is the same as arm_nn_conv1d_k3_nhwc_f16.

Parameters

Parameters of arm_nn_conv1d_k3_nhwc_f16_acc16

Name	Type	Direction	Description
x_nhwc	const float16_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of input channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
kernel	const float16_t *	in	Weights with shape `[out_c][3][in_c]`.
b	const float16_t *	in	Optional bias vector of `out_c` elements. May be NULL.
out	float16_t *	out	Output row in NHWC layout with shape `[out_w][out_c]`.
out_c	int32_t	in	Number of output channels.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `ow..ow+2`.

function

arm_nn_conv1d_k3_packed_f16

C

Specialized NHWC 1D convolution kernel for k=3 (float16, packed weights).

Include/arm_nnsupportfunctions_flt.h:1218

void arm_nn_conv1d_k3_packed_f16(
    const float16_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    const float16_t *kernel_packed,
    const float16_t *b,
    float16_t *out,
    int32_t out_c,
    int32_t out_w
)

Specialized NHWC 1D convolution kernel for k=3 (float16, packed weights).

The packed kernel uses the same NTxN RHS layout as arm_nn_mat_mult_nt_n_packed_f16, i.e. [(3 * in_c)][out_c_block_of_8].

Note

Accumulation width per leg. Scalar leg (non-MVE builds and ARM_MATH_AUTOVECTORIZE): bias and products accumulate in float32 and round to float16 once at the store (AmbiqAI/ns-cmsis-nn#449, #465). MVE leg: blockwise (#586): a lane’s float16 partial covers at most 10 input channels (30 taps, bias first) before it is widened into a float32 accumulator; one rounding at the store. Up to 32 taps (in_c <= 10) this is the float16-lane result.

Parameters

Parameters of arm_nn_conv1d_k3_packed_f16

Name	Type	Direction	Description
x_nhwc	const float16_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of input channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
kernel_packed	const float16_t *	in	Weights packed in output-channel blocks of 8 as described above.
b	const float16_t *	in	Optional bias vector of `out_c` elements. May be NULL.
out	float16_t *	out	Output row in NHWC layout with shape `[out_w][out_c]`.
out_c	int32_t	in	Number of output channels.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `ow..ow+2`.

function

arm_nn_conv1d_k3_packed_f16_acc16

C

Specialized NHWC 1D convolution kernel for k=3 (float16, packed weights).

Include/arm_nnsupportfunctions_flt.h:1233

void arm_nn_conv1d_k3_packed_f16_acc16(
    const float16_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    const float16_t *kernel_packed,
    const float16_t *b,
    float16_t *out,
    int32_t out_c,
    int32_t out_w
)

Specialized NHWC 1D convolution kernel for k=3 (float16, packed weights).

The packed kernel uses the same NTxN RHS layout as arm_nn_mat_mult_nt_n_packed_f16, i.e. [(3 * in_c)][out_c_block_of_8].

Note

Accumulation width per leg. Scalar leg (non-MVE builds and ARM_MATH_AUTOVECTORIZE): bias and products accumulate in float32 and round to float16 once at the store (AmbiqAI/ns-cmsis-nn#449, #465). MVE leg: blockwise (#586): a lane’s float16 partial covers at most 10 input channels (30 taps, bias first) before it is widened into a float32 accumulator; one rounding at the store. Up to 32 taps (in_c <= 10) this is the float16-lane result.

Note

Every MVE accumulator lane stays in float16 (no blockwise fold, AmbiqAI/ns-cmsis-nn#586); the scalar leg is the same as arm_nn_conv1d_k3_packed_f16.

Parameters

Parameters of arm_nn_conv1d_k3_packed_f16_acc16

Name	Type	Direction	Description
x_nhwc	const float16_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of input channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
kernel_packed	const float16_t *	in	Weights packed in output-channel blocks of 8 as described above.
b	const float16_t *	in	Optional bias vector of `out_c` elements. May be NULL.
out	float16_t *	out	Output row in NHWC layout with shape `[out_w][out_c]`.
out_c	int32_t	in	Number of output channels.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `ow..ow+2`.

function

arm_nn_maxpool1d_k3s3_nhwc_f16

C

Specialized NHWC max-pool 1D kernel for k=3, s=3 (float16).

Include/arm_nnsupportfunctions_flt.h:1247

void arm_nn_maxpool1d_k3s3_nhwc_f16(const float16_t *x_nhwc, int32_t in_c, int32_t in_w, float16_t *out, int32_t out_w)

Specialized NHWC max-pool 1D kernel for k=3, s=3 (float16).

Parameters

Parameters of arm_nn_maxpool1d_k3s3_nhwc_f16

Name	Type	Direction	Description
x_nhwc	const float16_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
out	float16_t *	out	Output row in NHWC layout with shape `[out_w][in_c]`.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `3*ow..3*ow+2`.

function

arm_nn_maxpool1d_k2s2_nhwc_noclip_f16

C

Specialized NHWC max-pool 1D kernel for k=2, s=2 without output clamp (float16).

Include/arm_nnsupportfunctions_flt.h:1258

void arm_nn_maxpool1d_k2s2_nhwc_noclip_f16(
    const float16_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    float16_t *out,
    int32_t out_w
)

Specialized NHWC max-pool 1D kernel for k=2, s=2 without output clamp (float16).

Parameters

Parameters of arm_nn_maxpool1d_k2s2_nhwc_noclip_f16

Name	Type	Direction	Description
x_nhwc	const float16_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
out	float16_t *	out	Output row in NHWC layout with shape `[out_w][in_c]`.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `2*ow..2*ow+1`.

function

arm_nn_maxpool1d_k2s2_nhwc_f16

C

Specialized NHWC max-pool 1D kernel for k=2, s=2 with clamp (float16).

Include/arm_nnsupportfunctions_flt.h:1269

void arm_nn_maxpool1d_k2s2_nhwc_f16(
    const float16_t *x_nhwc,
    int32_t in_c,
    int32_t in_w,
    float16_t *out,
    int32_t out_w,
    float16_t act_min,
    float16_t act_max
)

Specialized NHWC max-pool 1D kernel for k=2, s=2 with clamp (float16).

Parameters

Parameters of arm_nn_maxpool1d_k2s2_nhwc_f16

Name	Type	Direction	Description
x_nhwc	const float16_t *	in	Input row in NHWC layout with shape `[in_w][in_c]`.
in_c	int32_t	in	Number of channels.
in_w	int32_t	in	Input width. Currently unused by the kernel.
out	float16_t *	out	Output row in NHWC layout with shape `[out_w][in_c]`.
out_w	int32_t	in	Output width. Output position `ow` reads input positions `2*ow..2*ow+1`.
act_min	float16_t	in	Lower clamp bound applied to `out`.
act_max	float16_t	in	Upper clamp bound applied to `out`.

function

arm_nn_mat_mult_nt_t_f16

C

Matrix multiply with non-transposed lhs and transposed rhs rows (float32).

Include/arm_nnsupportfunctions_flt.h:1293

arm_cmsis_nn_status arm_nn_mat_mult_nt_t_f16(
    const float16_t *lhs,
    const float16_t *rhs,
    const float16_t *bias,
    float16_t *dst,
    int32_t lhs_rows,
    int32_t rhs_rows,
    int32_t rhs_cols,
    int32_t row_address_offset,
    float16_t activation_min,
    float16_t activation_max
)

Matrix multiply with non-transposed lhs and transposed rhs rows (float32).

Note

Accumulation width per leg. MVE legs accumulate blockwise (AmbiqAI/ns-cmsis-nn#586, superseding the float16-lane choice of #417 / #446 for the MVE legs). Up to rhs_cols 32 nothing changes: per-k float16 lanes on the gather path (rhs_cols below the contiguous-K threshold), lane-partial sums then one float16 reduction plus the bias in float16 at rhs_cols 32. Above 32, on the contiguous-K path and the remainder rows, each lane (element k goes to lane k % 8) sums at most 32 of its own taps (256 elements) in float16; each block’s lanes are then widened and lanes 2j and 2j+1 added in float32 into pair accumulator j (set by the first block, added to by later ones); the four pair accumulators are summed once as (0+1) + (2+3), so a single block sums ((0+1) + (2+3)) + ((4+5) + (6+7)), the bias is added in float32 and the total rounds to float16 once before the clamp. The float16 reduction up to rhs_cols 32 is ordered by the compiler, which may reorder it under -ffast-math; only the fold’s order is fixed. arm_nn_mat_mult_nt_t_f16_acc16 keeps the float16 lanes and float16 reduction throughout. The scalar leg (non-MVE builds and ARM_MATH_AUTOVECTORIZE) accumulates bias and every product in float32 and rounds to float16 once before the clamp (AmbiqAI/ns-cmsis-nn#449, #457).

Parameters

Parameters of arm_nn_mat_mult_nt_t_f16

Name	Type	Direction	Description
lhs	const float16_t *	in	Left-hand matrix stored row-major.
rhs	const float16_t *	in	Right-hand matrix stored row-major, one row per output channel.
bias	const float16_t *	in	Optional bias vector.
dst	float16_t *	out	Output matrix.
lhs_rows	int32_t	in	Number of rows in `lhs`.
rhs_rows	int32_t	in	Number of rows in `rhs`.
rhs_cols	int32_t	in	Number of columns in `rhs`.
row_address_offset	int32_t	in	Output row stride, expressed in elements.
activation_min	float16_t	in	Lower clamp bound.
activation_max	float16_t	in	Upper clamp bound.

Returns

Returns of arm_nn_mat_mult_nt_t_f16

Description
`ARM_CMSIS_NN_SUCCESS` on success or `ARM_CMSIS_NN_ARG_ERROR` on invalid arguments.

function

arm_nn_mat_mult_nt_t_f16_acc16

C

armnnmatmultnttf16 with every MVE accumulator lane in float16 (no blockwise fold).

Include/arm_nnsupportfunctions_flt.h:1321

arm_cmsis_nn_status arm_nn_mat_mult_nt_t_f16_acc16(
    const float16_t *lhs,
    const float16_t *rhs,
    const float16_t *bias,
    float16_t *dst,
    int32_t lhs_rows,
    int32_t rhs_rows,
    int32_t rhs_cols,
    int32_t row_address_offset,
    float16_t activation_min,
    float16_t activation_max
)

arm_nn_mat_mult_nt_t_f16 with every MVE accumulator lane in float16 (no blockwise fold).

Same arguments, return codes and scalar leg as arm_nn_mat_mult_nt_t_f16; see its accumulation note.

Parameters

Parameters of arm_nn_mat_mult_nt_t_f16_acc16

Name	Type	Direction	Description
lhs	const float16_t *	in	Left-hand matrix, row-major `[lhs_rows, rhs_cols]`.
rhs	const float16_t *	in	Right-hand matrix, row-major `[rhs_rows, rhs_cols]` (transposed operand).
bias	const float16_t *	in	Optional bias vector of `rhs_rows` elements.
dst	float16_t *	out	Output matrix.
lhs_rows	int32_t	in	Number of rows in `lhs`.
rhs_rows	int32_t	in	Number of rows in `rhs`.
rhs_cols	int32_t	in	Shared reduction dimension `K`.
row_address_offset	int32_t	in	Output row stride, expressed in elements.
activation_min	float16_t	in	Lower clamp bound.
activation_max	float16_t	in	Upper clamp bound.

Returns

Returns of arm_nn_mat_mult_nt_t_f16_acc16

Description
`ARM_CMSIS_NN_SUCCESS` on success or `ARM_CMSIS_NN_ARG_ERROR` on invalid arguments.

function

arm_nn_mat_mult_nt_n_packed_f16

C

Matrix multiply with non-transposed lhs and packed non-transposed rhs (float16).

Include/arm_nnsupportfunctions_flt.h:1362

arm_cmsis_nn_status arm_nn_mat_mult_nt_n_packed_f16(
    const float16_t *lhs,
    const float16_t *rhs_packed,
    const float16_t *bias,
    float16_t *dst,
    int32_t lhs_rows,
    int32_t rhs_rows,
    int32_t rhs_cols,
    int32_t row_address_offset,
    float16_t activation_min,
    float16_t activation_max
)

Matrix multiply with non-transposed lhs and packed non-transposed rhs (float16).

Note

On non-MVE builds the output clamp is the bit-classified scalar clamp of #380, so a NaN accumulator (a NaN in lhs, rhs_packed or bias) propagates to dst at every optimization level on the gated toolchains, including the shipped -Ofast. On MVE builds the clamp is vmaxnmq/vminnmq with no NaN restore, so a NaN resolves to a clamp bound there instead.

Note

Accumulation width per leg: the MVE leg accumulates blockwise (AmbiqAI/ns-cmsis-nn#586): one lane per output column, per-k, the bias opening the first block; every 32 k the float16 partial is widened exactly into per-lane float32 accumulators, which round to float16 once before the clamp (rhs_cols up to 32: exactly the float16-lane result; arm_nn_mat_mult_nt_n_packed_f16_acc16 keeps float16 lanes throughout). The scalar leg (non-MVE builds and ARM_MATH_AUTOVECTORIZE) accumulates bias and every product in float32 and rounds to float16 once before the clamp (AmbiqAI/ns-cmsis-nn#449, #457).

Parameters

Parameters of arm_nn_mat_mult_nt_n_packed_f16

Name	Type	Direction	Description
lhs	const float16_t *	in	Left-hand matrix stored row-major with logical shape `[lhs_rows, rhs_cols]`.
rhs_packed	const float16_t *	in	Right-hand matrix with logical shape `[rhs_cols, rhs_rows]`, packed in column blocks of 8. The final block uses the same packed stride and inactive tail lanes are ignored.
bias	const float16_t *	in	Optional bias vector.
dst	float16_t *	out	Output matrix.
lhs_rows	int32_t	in	Number of rows in `lhs`.
rhs_rows	int32_t	in	Number of logical output columns in the unpacked rhs matrix.
rhs_cols	int32_t	in	Shared reduction dimension `K`.
row_address_offset	int32_t	in	Output row stride, expressed in elements.
activation_min	float16_t	in	Lower clamp bound.
activation_max	float16_t	in	Upper clamp bound.

Returns

Returns of arm_nn_mat_mult_nt_n_packed_f16

Description
`ARM_CMSIS_NN_SUCCESS` on success or `ARM_CMSIS_NN_ARG_ERROR` on invalid arguments.

function

arm_nn_mat_mult_nt_n_packed_f16_precise

C

Packed float16 matrix multiply with shorter partial sums, for precise packed callers.

Include/arm_nnsupportfunctions_flt.h:1399

arm_cmsis_nn_status arm_nn_mat_mult_nt_n_packed_f16_precise(
    const float16_t *lhs,
    const float16_t *rhs_packed,
    const float16_t *bias,
    float16_t *dst,
    int32_t lhs_rows,
    int32_t rhs_rows,
    int32_t rhs_cols,
    int32_t row_address_offset,
    float16_t activation_min,
    float16_t activation_max
)

Packed float16 matrix multiply with shorter partial sums, for precise packed callers.

The packed layout and all argument meanings are those of arm_nn_mat_mult_nt_n_packed_f16. This entry has its own numerical order; it does not change that entry or its acc16 twin.

Note

MVE: each lane forms at most four fused-product taps in float16, starting from zero. Each partial is widened and added to its float32 lane accumulator in K order; bias is added in float32 after the final fold, then the result rounds to float16 once before clamping. The scalar/autovectorized leg uses the existing float32 accumulation rule. This order is not bit-equivalent to the standard or legacy packed entries, and no universal absolute-error bound is implied. Float16 products/partials can overflow for finite inputs; shorter partials do not make this an overflow-free float32 dot product. As with the legacy packed entry, the scalar clamp propagates NaN, while the MVE maxNum/minNum clamp resolves NaN to a bound.

Parameters

Parameters of arm_nn_mat_mult_nt_n_packed_f16_precise

Name	Type	Direction	Description
lhs	const float16_t *	in	Non-transposed lhs, with lhs_rows rows of rhs_cols elements.
rhs_packed	const float16_t *	in	Logical rhs_rows output columns in 8-column blocks, K-major within each block.
bias	const float16_t *	in	Optional rhs_rows-element bias; NULL means zero bias.
dst	float16_t *	out	Output matrix. The caller provides the complete arrays, including packed column padding.
lhs_rows	int32_t	in	Positive number of lhs rows.
rhs_rows	int32_t	in	Positive number of logical output columns.
rhs_cols	int32_t	in	Positive shared reduction dimension K.
row_address_offset	int32_t	in	Positive output row stride, in elements.
activation_min	float16_t	in	Lower ordered clamp bound.
activation_max	float16_t	in	Upper ordered clamp bound.

Returns

Returns of arm_nn_mat_mult_nt_n_packed_f16_precise

Description
ARM_CMSIS_NN_SUCCESS, or ARM_CMSIS_NN_ARG_ERROR for NULL lhs/rhs/dst or nonpositive dimensions/stride.

function

arm_nn_mat_mult_nt_n_packed_f16_acc16

C

armnnmatmultntnpackedf16 with every MVE accumulator lane in float16 (no blockwise fold).

Include/arm_nnsupportfunctions_flt.h:1427

arm_cmsis_nn_status arm_nn_mat_mult_nt_n_packed_f16_acc16(
    const float16_t *lhs,
    const float16_t *rhs_packed,
    const float16_t *bias,
    float16_t *dst,
    int32_t lhs_rows,
    int32_t rhs_rows,
    int32_t rhs_cols,
    int32_t row_address_offset,
    float16_t activation_min,
    float16_t activation_max
)

arm_nn_mat_mult_nt_n_packed_f16 with every MVE accumulator lane in float16 (no blockwise fold).

Same arguments, return codes and scalar leg as arm_nn_mat_mult_nt_n_packed_f16; see its accumulation note.

Parameters

Parameters of arm_nn_mat_mult_nt_n_packed_f16_acc16

Name	Type	Direction	Description
lhs	const float16_t *	in	Left-hand matrix stored row-major with logical shape `[lhs_rows, rhs_cols]`.
rhs_packed	const float16_t *	in	Right-hand matrix packed in column blocks of 8.
bias	const float16_t *	in	Optional bias vector.
dst	float16_t *	out	Output matrix.
lhs_rows	int32_t	in	Number of rows in `lhs`.
rhs_rows	int32_t	in	Number of logical output columns in the unpacked rhs matrix.
rhs_cols	int32_t	in	Shared reduction dimension `K`.
row_address_offset	int32_t	in	Output row stride, expressed in elements.
activation_min	float16_t	in	Lower clamp bound.
activation_max	float16_t	in	Upper clamp bound.

Returns

Returns of arm_nn_mat_mult_nt_n_packed_f16_acc16

Description
`ARM_CMSIS_NN_SUCCESS` on success or `ARM_CMSIS_NN_ARG_ERROR` on invalid arguments.

function

arm_nn_lstm_step_f16

C

Update LSTM function for an iteration step using float16 input, output and state.

Include/arm_nnsupportfunctions_flt.h:1451

arm_cmsis_nn_status arm_nn_lstm_step_f16(
    const float16_t *data_in,
    const float16_t *hidden_in,
    float16_t *hidden_out,
    const cmsis_nn_lstm_params_f16 *params,
    cmsis_nn_lstm_context_f16 *buffers,
    const int32_t batch_offset
)

Update LSTM function for an iteration step using float16 input, output and state.

Parameters

Parameters of arm_nn_lstm_step_f16

Name	Type	Direction	Description
data_in	const float16_t *	in	Data input pointer.
hidden_in	const float16_t *	in	Hidden state / recurrent input pointer. May be NULL for the first step.
hidden_out	float16_t *	out	Hidden state / recurrent output pointer.
params	const cmsis_nn_lstm_params_f16 *	in	Struct containing all information about the LSTM operator.
buffers	cmsis_nn_lstm_context_f16 *	in, out	Struct containing pointers to mutable cell-state storage.
batch_offset	const int32_t	in	Number of timesteps between consecutive batches.

Returns

Returns of arm_nn_lstm_step_f16

Description
ARM_CMSIS_NN_SUCCESS on success, or ARM_CMSIS_NN_ARG_ERROR on invalid arguments (NULL data_in/hidden_out/params/buffers or buffers->cell_state, batch_offset <= 0).

function

arm_nn_gru_step_f16

C

Update GRU function for a single iteration step using float16 data.

Include/arm_nnsupportfunctions_flt.h:1471

arm_cmsis_nn_status arm_nn_gru_step_f16(
    const float16_t *data_in,
    const float16_t *hidden_in,
    float16_t *hidden_out,
    const cmsis_nn_gru_params_f16 *params,
    cmsis_nn_gru_context_f16 *buffers,
    const int32_t batch_offset
)

Update GRU function for a single iteration step using float16 data.

Parameters

Parameters of arm_nn_gru_step_f16

Name	Type	Direction	Description
data_in	const float16_t *	in	Data input pointer for this time step.
hidden_in	const float16_t *	in	Recurrent input pointer. NULL for the first step (h_prev = 0).
hidden_out	float16_t *	out	Hidden-state output pointer for this time step.
params	const cmsis_nn_gru_params_f16 *	in	Struct describing the GRU operator.
buffers	cmsis_nn_gru_context_f16 *	in, out	Scratch buffers. temp1 (>= hidden_size) is required when reset_after == 0.
batch_offset	const int32_t	in	Number of timesteps between consecutive batches.

Returns

Returns of arm_nn_gru_step_f16

Description
ARM_CMSIS_NN_SUCCESS on success, or ARM_CMSIS_NN_ARG_ERROR on invalid arguments (NULL data_in/hidden_out/params, batch_offset <= 0, or missing temp1 when reset_after == 0).

function

arm_nn_pack_conv_patch_f16

C

Pack a single convolution patch into one row of a contiguous float32 patch matrix.

Include/arm_nnsupportfunctions_flt.h:1481

void arm_nn_pack_conv_patch_f16(
    const float16_t *input,
    int32_t in_h,
    int32_t in_w,
    int32_t in_c,
    int32_t kernel_h,
    int32_t kernel_w,
    int32_t stride_h,
    int32_t stride_w,
    int32_t pad_h,
    int32_t pad_w,
    int32_t dilation_h,
    int32_t dilation_w,
    int32_t out_y,
    int32_t out_x,
    float16_t pad_value,
    float16_t *patch_row
)

Pack a single convolution patch into one row of a contiguous float32 patch matrix.

Developers familiar with im2row/im2col terminology can think of this as packing one output patch into one row.

Parameters

Parameters of arm_nn_pack_conv_patch_f16

Name	Type	Direction	Description
input	const float16_t *	in	Input tensor for one batch in NHWC layout with shape `[in_h][in_w][in_c]`.
in_h	int32_t	in	Input height.
in_w	int32_t	in	Input width.
in_c	int32_t	in	Number of input channels.
kernel_h	int32_t	in	Kernel height.
kernel_w	int32_t	in	Kernel width.
stride_h	int32_t	in	Vertical stride.
stride_w	int32_t	in	Horizontal stride.
pad_h	int32_t	in	Top padding.
pad_w	int32_t	in	Left padding.
dilation_h	int32_t	in	Vertical dilation.
dilation_w	int32_t	in	Horizontal dilation.
out_y	int32_t	in	Output row index of the patch to pack.
out_x	int32_t	in	Output column index of the patch to pack.
pad_value	float16_t	in	Value written for taps that fall outside the input.
patch_row	float16_t *	out	Destination row of `kernel_h * kernel_w * in_c` elements, ordered `[kernel_h][kernel_w][in_c]`.

function

arm_nn_softmax_1x2_f16

C

Specialized softmax helper for a single float16 row of length 2.

Include/arm_nnsupportfunctions_flt.h:1504

void arm_nn_softmax_1x2_f16(const float16_t *in, float16_t *out)

Specialized softmax helper for a single float16 row of length 2.

Parameters

Parameters of arm_nn_softmax_1x2_f16

Name	Type	Direction	Description
in	const float16_t *	in	Pointer to two contiguous float16 input values.
out	float16_t *	out	Pointer to two contiguous float16 output values.

function

arm_nn_lstm_step_f32

C

Update LSTM function for an iteration step using float32 input, output and state.

Include/arm_nnsupportfunctions_flt.h:1523

arm_cmsis_nn_status arm_nn_lstm_step_f32(
    const float32_t *data_in,
    const float32_t *hidden_in,
    float32_t *hidden_out,
    const cmsis_nn_lstm_params_f32 *params,
    cmsis_nn_lstm_context_f32 *buffers,
    const int32_t batch_offset
)

Update LSTM function for an iteration step using float32 input, output and state.

Parameters

Parameters of arm_nn_lstm_step_f32

Name	Type	Direction	Description
data_in	const float32_t *	in	Data input pointer.
hidden_in	const float32_t *	in	Hidden state / recurrent input pointer. May be NULL for the first step.
hidden_out	float32_t *	out	Hidden state / recurrent output pointer.
params	const cmsis_nn_lstm_params_f32 *	in	Struct containing all information about the LSTM operator.
buffers	cmsis_nn_lstm_context_f32 *	in, out	Struct containing pointers to mutable cell-state storage.
batch_offset	const int32_t	in	Number of timesteps between consecutive batches.

Returns

Returns of arm_nn_lstm_step_f32

Description
ARM_CMSIS_NN_SUCCESS on success, or ARM_CMSIS_NN_ARG_ERROR on invalid arguments (NULL data_in/hidden_out/params/buffers or buffers->cell_state, batch_offset <= 0).

function

arm_nn_gru_step_f32

C

Update GRU function for a single iteration step using float32 data.

Include/arm_nnsupportfunctions_flt.h:1543

arm_cmsis_nn_status arm_nn_gru_step_f32(
    const float32_t *data_in,
    const float32_t *hidden_in,
    float32_t *hidden_out,
    const cmsis_nn_gru_params_f32 *params,
    cmsis_nn_gru_context_f32 *buffers,
    const int32_t batch_offset
)

Update GRU function for a single iteration step using float32 data.

Parameters

Parameters of arm_nn_gru_step_f32

Name	Type	Direction	Description
data_in	const float32_t *	in	Data input pointer for this time step.
hidden_in	const float32_t *	in	Recurrent input pointer. NULL for the first step (h_prev = 0).
hidden_out	float32_t *	out	Hidden-state output pointer for this time step.
params	const cmsis_nn_gru_params_f32 *	in	Struct describing the GRU operator.
buffers	cmsis_nn_gru_context_f32 *	in, out	Scratch buffers. temp1 (>= hidden_size) is required when reset_after == 0.
batch_offset	const int32_t	in	Number of timesteps between consecutive batches.

Returns

Returns of arm_nn_gru_step_f32

Description
ARM_CMSIS_NN_SUCCESS on success, or ARM_CMSIS_NN_ARG_ERROR on invalid arguments (NULL data_in/hidden_out/params, batch_offset <= 0, or missing temp1 when reset_after == 0).